CN108460032A - A kind of generation method and device of video frequency abstract - Google Patents

Abstract

An embodiment of the present invention provides a kind of generation method of video frequency abstract and devices, wherein the generation method of video frequency abstract includes：Target retrieval condition is obtained, established database is retrieved, obtains the first track set comprising all tracks for meeting target retrieval condition；Overlapping at least two tracks occur according to overlapping status information, during the first track is gathered and are divided into one group, every group is determined as a combined trajectories；It will meet the trajectory-offset of presetting translation condition in track to be translated to same target time section along time shaft；From the corresponding all first object artworks in all tracks obtained in database in target time section；Obtain the first abstract Background for generating video frequency abstract；All first object artworks and the first abstract Background are spliced, video frequency abstract is generated.The visual effect of the video frequency abstract generated in the case of target trajectory complexity can be improved through the invention.

Description

A kind of generation method and device of video frequency abstract

Technical field

The present invention relates to technical field of video processing, more particularly to the generation method and device of a kind of video frequency abstract.

Background technology

With to video data processing requirement continuous improvement and the video data volume it is increasingly huge, user wants to Section video is grown for one and establishes an abstract, by fast browsing in order to preferably utilize the video, such as applied to maintenance The video monitoring of social security, the delinquent phenomenon of strike.Therefore, video summarization technique is come into being.Video summarization technique is The structure and content of video are analyzed, extract significant part, i.e. moving target from original video, and by the movement Target is combined in a specific way with background scene, forms the succinct summary that can fully show video content；Video is plucked If to the simplified summary of long video content, usually indicated with one section of image sequence either statically or dynamically, and to raw information Retained.

In the prior art, the technology for generating the most mainstream of video frequency abstract is the generation skill of the video frequency abstract based on target object Art, the technology comprise the following steps：First, by the analysis to input video, video structural description file is generated, and according to The video structural description file establishes Relational database, wherein includes the mesh in video in video structural description file Mark the trace information of the attribute information and target object of object；Secondly, the database of foundation is retrieved, extracts moving target Trace information；Finally, then to the track of all target objects it analyzes, translates target trajectory on a timeline, it will be different In the track arrangement to same picture of the target object of time, summarized radio is generated with this.The technology disclosure satisfy that user to spy The object that sets the goal generate video frequency abstract demand, and generate video frequency abstract when length, movement it is compact, have higher concentration Than.

But when the track of target object is arranged, actual conditions are complicated, overlapped there are the track of multiple target objects The track of overlapping target object will occur for situation, the method due to above-mentioned technology by rearranging the track of target object It is excluded, therefore when the track of multiple target objects occurs overlapping, summarized radio can be caused to lose the letter of associated objects object Breath；And due to the loss of associated objects object information so that a target pair can be frequently occurred in the video frequency abstract of generation The phenomenon that as generating and disappearing suddenly, causes visual effect bad.

Invention content

The embodiment of the present invention is designed to provide a kind of generation method and device of video frequency abstract, to improve in target track The visual effect of the video frequency abstract generated in the case of mark complexity.Specific technical solution is as follows：

In a first aspect, an embodiment of the present invention provides a kind of generation method of video frequency abstract, the method includes：

Target retrieval condition is obtained, established database is retrieved, obtains including to meet the target retrieval item Gather first track of all tracks of part, wherein be stored in the database and carried from the video frame comprising target object The trace information and target artwork of the every track taken, the trace information of every track include：Between other tracks Overlapping status information；

Overlapping at least two tracks occur according to the overlapping status information, during first track is gathered to be divided into One group, every group is determined as a combined trajectories；

It will meet the trajectory-offset of presetting translation condition in track to be translated to same target time section along time shaft, In, the track to be translated includes：Overlapping track does not occur in the combined trajectories and/or first track set；

The corresponding all first object artworks in all tracks in the target time section are obtained from the database；

Obtain the first abstract Background for generating video frequency abstract；

All first object artworks and the first abstract Background are spliced, video frequency abstract is generated.

Optionally, the target retrieval condition includes：The searching attribute information of retrieval time section and/or target object；

It is described that established database is retrieved when search condition only includes retrieval time section, including：

According to retrieval time section, the database is retrieved, obtains all targets in the retrieval time section The track of object；

It is described that established database is examined when search condition only includes the searching attribute information of target object Rope, including：

According to the searching attribute information of the target object, the database is retrieved, obtains and belongs to the retrieval All tracks of all target objects of property information matches；

It is described to established data when search condition includes the searching attribute information of retrieval time section and target object Library is retrieved, including：

According to the searching attribute information of retrieval time section and the searched targets object, the database is examined Rope obtains in the retrieval time section, all tracks with the matched all target objects of the searching attribute.

Optionally, before the acquisition target retrieval condition, the method further includes：

All target objects are extracted from the video of input；

Extract the trace information and attribute information of each target object, wherein trace information includes：The mobile letter of track Breath and the overlapping status information between other tracks；

The trace information of each target object and attribute information are stored into video structural goal description file；

According to the video structural goal description file, the database is generated；

From the video frame comprising target object, the artwork and mask figure of the corresponding each frame of extraction trace information, according to The artwork and the mask figure determine the target artwork of each frame corresponding with the trace information；

The artwork, the mask figure and the target artwork are stored into the database.

Optionally, described from the video frame comprising target object, the mask figure of the corresponding each frame of extraction trace information, Including：

From the video frame comprising target object, the motion mask of the target object is extracted；

According to the motion mask, initial mask figure is determined；

Determine the edge point set of the initial mask figure；

The convex set that the marginal point is concentrated is extracted, the convex closure point set of the mask figure is constituted；

The corresponding convex closure of the convex closure point set is filled, final mask figure is obtained.

Optionally, described to meet the trajectory-offset of presetting translation condition in track to be translated to same target along time shaft Period, including：

Establish queue to be translated and abstract queue；

Using first track gather in not the track in the target time section as track to be translated, and store to In queue to be translated, wherein the track to be translated be first track set in not in the target time section Combined trajectories and overlapping track does not occur；

The track in the target time section during first track is gathered is stored into the abstract queue；

Current track to be translated is extracted from the queue to be translated successively, and according to the artwork in the database, is obtained To the rectangle frame of all target objects in the corresponding video frame in the current track to be translated；

Calculate the rectangle frames of all target objects respectively with stored it is corresponding to abstract every track of queue Video frame in target object rectangle frame between overlapping area；

When the overlapping area is less than or equal to default Overlapping parameters threshold value, extremely by the current trajectory-offset to be translated The target time section, and store to the abstract queue；

The corresponding all first object artworks in all tracks for obtaining the target time section from the database The step of, including：

Obtain the corresponding all first object artworks in all tracks in the abstract queue.

Optionally, further include in the trace information of every track：The target frame information aggregate of the target object；

It is described to splice all first object artworks and the first abstract Background, video frequency abstract is generated, Including：

According to the target frame information aggregate of target object described in the trace information, determine that the first object artwork exists First position in the first abstract Background；

Each first object artwork is copied into corresponding first position in the first abstract Background, generation regards Frequency is made a summary.

Optionally, described to copy to each first object artwork corresponding first in the first abstract Background Position, including：

If there is overlapping target object in each first object artwork, the pixel of the overlapping part of corresponding track is set Value is the mean value of the target artwork pixel value of each target object, the pixel value of overlapping part is not the target artwork of each target object Pixel value, obtain figure to be copied；

It is described that each first object artwork is copied into corresponding first position in the first abstract Background, it is raw At video frequency abstract, including：

Each first object artwork that each figure and target object to be copied do not overlap is copied to described first Corresponding first position in abstract Background, generates video frequency abstract.

Optionally, before storing the artwork, the mask figure and the target artwork into the database, institute The method of stating further includes：

By predetermined period, abstract Background is obtained；

The abstract Background in each period of acquisition is stored into the database；

The first abstract Background obtained for generating video frequency abstract, including：

By the time corresponding to each predetermined period, the target time section is divided into corresponding time subsegment, Wherein, a time subsegment corresponds to a predetermined period；

It determines in the target time section, comprising corresponding first predetermined period of the most time subsegment in track；

From the database, the corresponding first abstract Background of first predetermined period is obtained.

Optionally, before the acquisition target retrieval condition, the method further includes：

According to user instruction, user interface is shown；

Receive and preserve target retrieval condition, default translation condition and use that user is inputted by the user interface In the predetermined period for generating abstract Background；

The method further includes：

When receiving the startup request that user is inputted by the user interface, execute described to established number The step of being retrieved according to library；

When receiving the interrupt requests that user is inputted by the user interface, terminate the stream that video frequency abstract generates Journey.

Second aspect, an embodiment of the present invention provides a kind of generating means of video frequency abstract, described device includes：

Retrieval module retrieves established database for obtaining target retrieval condition, obtains including to meet institute State the first track set of all tracks of target retrieval condition, wherein be stored in the database from comprising target object Video frame in the trace information and target artwork of the every track that extract, the trace information of every track includes：With Overlapping status information between other tracks；

Composite module, for according to the overlapping status information, occurring to overlap at least during first track is gathered Two tracks are divided into one group, and every group is determined as a combined trajectories；

Translation module, for the trajectory-offset for presetting translation condition will to be met in track to be translated to same mesh along time shaft Mark the period, wherein the track to be translated includes：It is not handed in the combined trajectories and/or first track set Folded track；

First acquisition module, for obtaining the corresponding institute in all tracks in the target time section from the database There is first object artwork；

Second acquisition module, for obtaining the first abstract Background for generating video frequency abstract；

Concatenation module is generated for splicing all first object artworks and the first abstract Background Video frequency abstract.

Optionally, the target retrieval condition includes：The searching attribute information of retrieval time section and/or target object；

When search condition only includes retrieval time section, the retrieval module is specifically used for：

According to retrieval time section, the database is retrieved, obtains all targets in the retrieval time section The track of object is as target trajectory；

When search condition only includes the searching attribute information of target object, the retrieval module is specifically used for：

According to the searching attribute information of the target object, the database is retrieved, obtains and belongs to the retrieval All tracks of all target objects of property information matches are as target trajectory；

When search condition includes the searching attribute information of retrieval time section and target object, the retrieval module, specifically For：

According to the searching attribute information of retrieval time section and the searched targets object, the database is examined Rope obtains in the retrieval time section, and all tracks with the matched all target objects of the searching attribute are as target track Mark.

Optionally, described device further includes：

First extraction module, for extracting all target objects from the video of input；

Second extraction module, trace information and attribute information for extracting each target object, wherein in trace information Including：The mobile message of track and the overlapping status information between other tracks；

First memory module, for storing the trace information of each target object and attribute information to video structure Change in goal description file；

Database generation module, for according to the video structural goal description file, generating the database；

Third extraction module, for from the video frame comprising target object, extracting the corresponding each frame of trace information Artwork and mask figure determine that the target of each frame corresponding with the trace information is former according to the artwork and the mask figure Figure；

Second memory module, for storing the artwork, the mask figure and the target artwork to the database In.

Optionally, the third extraction module, including：

First extracting sub-module, for from the video frame comprising target object, the movement for extracting the target object to be covered Code；

First determination sub-module, for according to the motion mask, determining initial mask figure；

Second determination sub-module, the edge point set for determining the initial mask figure；

Second extracting sub-module, the convex set concentrated for extracting the marginal point, constitutes the convex closure point set of the mask figure；

Filling submodule obtains final mask figure for filling the corresponding convex closure of the convex closure point set.

Optionally, the translation module, including：

Queue setting up submodule, for establishing queue to be translated and abstract queue；

First sub-module stored, for using first track gather in not the track in the target time section as Track to be translated, and store into queue to be translated, wherein the track to be translated is not existing in the set of first track Combined trajectories in the target time section and overlapping track does not occur；

Second sub-module stored is stored for the track in the target time section in gathering first track To in the abstract queue；

Third extracting sub-module, for extracting current track to be translated from the queue to be translated successively, and according to institute The artwork in database is stated, the rectangle frame of all target objects in the corresponding video frame in the track currently to be translated is obtained；

Operation submodule, for calculate the rectangle frames of all target objects respectively with stored to the abstract queue The corresponding video frame in every track in target object rectangle frame between overlapping area；

Third sub-module stored is used for when the overlapping area is less than or equal to default Overlapping parameters threshold value, will be described Current trajectory-offset to be translated is stored to the target time section to the abstract queue；

First acquisition module, is specifically used for：

Obtain the corresponding all first object artworks in all tracks in the abstract queue.

Optionally, further include in the trace information of every track：The target frame information aggregate of the target object；

The concatenation module, including：

Third determination sub-module, for the target frame information aggregate according to target object described in the trace information, really First position of the fixed first object artwork in the first abstract Background；

Video frequency abstract generates submodule, for each first object artwork to be copied to the first abstract Background In corresponding first position, generate video frequency abstract.

Optionally, the third determination sub-module, is specifically used for：

If there is overlapping target object in each first object artwork, the pixel of the overlapping part of corresponding track is set Value is the mean value of the target artwork pixel value of each target object, the pixel value of overlapping part is not the target artwork of each target object Pixel value, obtain figure to be copied；

The concatenation module, is specifically used for：

Each first object artwork that each figure and target object to be copied are not overlapped copies to described Corresponding first position in first abstract Background, generates video frequency abstract.

Optionally, described device further includes：

Computing module obtains abstract Background for pressing predetermined period；

Third memory module, for storing the abstract Background in each period obtained into the database；

Second acquisition module, including：

Submodule is divided, for by the time corresponding to each predetermined period, the target time section to be divided into Corresponding time subsegment, wherein a time subsegment corresponds to a predetermined period；

4th determination sub-module, for determining in the target time section, corresponding comprising the most time subsegment in track First predetermined period；

Background acquisition submodule is plucked for from the database, obtaining first predetermined period corresponding first Want Background.

Optionally, described device further includes：

Display module, for according to user instruction, showing user interface；

Receiving module, for receiving and preserving target retrieval condition that user inputted by the user interface, pre- If translation condition and the predetermined period for generating abstract Background；

Execution module, for when receiving the startup request that user is inputted by the user interface, executing institute State the step of being retrieved to established database；

Terminate module, for when receiving the interrupt requests that user is inputted by the user interface, terminating to regard The flow of frequency summarization generation.

The generation method and device of a kind of video frequency abstract provided in an embodiment of the present invention, by established database into Row retrieval, obtains the track for meeting target retrieval condition, and it is combined trajectories that overlapping track combination will occur in these tracks, so Combined trajectories in different time periods are translated afterwards and overlapping track does not occur to same target time section, finally to target time section In the corresponding target artwork in track and abstract Background spliced, generate video frequency abstract；The embodiment of the present invention is regarded in generation When frequency is made a summary, the track combinations of overlapping multiple target objects will occur into a combined trajectories, integral translation on a timeline, Certain tracks in losing overlapping track in translation are avoided, the visual effect for generating video frequency abstract is improved.

Description of the drawings

In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below There is attached drawing needed in technology description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this Some embodiments of invention for those of ordinary skill in the art without creative efforts, can be with Obtain other attached drawings according to these attached drawings.

Fig. 1 is the first flow diagram of the generation method of the video frequency abstract of the embodiment of the present invention；

Fig. 2 is second of flow diagram of the generation method of the video frequency abstract of the embodiment of the present invention；

Fig. 3 is the idiographic flow schematic diagram of S205 in embodiment illustrated in fig. 2；

Fig. 4 is the idiographic flow schematic diagram of S105 in embodiment illustrated in fig. 2；

Fig. 5 is the idiographic flow schematic diagram of S103 in embodiment illustrated in fig. 2；

Fig. 6 is the third flow diagram of the generation method of the video frequency abstract of the embodiment of the present invention；

Fig. 7 is the first structural schematic diagram of the generating means of the video frequency abstract of the embodiment of the present invention；

Fig. 8 is second of structural schematic diagram of the generating means of the video frequency abstract of the embodiment of the present invention；

Fig. 9 is the concrete structure schematic diagram of third extraction module 850 in embodiment illustrated in fig. 8；

Figure 10 is the concrete structure schematic diagram of translation module 730 in embodiment illustrated in fig. 8；

Figure 11 is the concrete structure schematic diagram of the second acquisition module 750 in embodiment illustrated in fig. 8；

Figure 12 is the third structural schematic diagram of the generating means of the video frequency abstract of the embodiment of the present invention.

Specific implementation mode

Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete Site preparation describes, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other Embodiment shall fall within the protection scope of the present invention.

In order to improve the visual effect of the video frequency abstract generated in the case of target trajectory complexity, the embodiment of the present invention carries A kind of generation method and device of video frequency abstract are supplied.

A kind of generation method of video frequency abstract is provided for the embodiments of the invention first below to be introduced.

It should be noted that a kind of executive agent of the generation method for video frequency abstract that the embodiment of the present invention is provided can be with There is the video frequency abstract controller for realizing generation video frequency abstract function to be a kind of.Wherein, realize what the embodiment of the present invention was provided A kind of mode of the generation method of video frequency abstract can be the software being set in video frequency abstract controller, hardware circuit and/or Logic circuit.Specifically, the video frequency abstract controller can be applied to video monitoring system, video website can also be applied to Server end.

As shown in Figure 1, a kind of generation method for video frequency abstract that the embodiment of the present invention is provided, may include walking as follows Suddenly：

S101 obtains target retrieval condition, is retrieved to established database, obtains including to meet target retrieval item Gather first track of all tracks of part.

Wherein, be stored in database the every track extracted from the video frame comprising target object trace information and Target artwork, the trace information of every track include the overlapping status information between other tracks；Database can also include： The mask figure and/or Background of target object in the attribute informations of all target objects, video frame artwork, each frame artwork；It hands over Overlapping state information can be the identifier for the track that overlapping target object occurs, and identifier can be the title of track, also may be used To be track number or other are used to characterize the characteristic symbol of track；Trace information may include：The track of target object Identifier, the tracing point quantity of target object, the frame number of each tracing point of target object, target object track time Information, target object track spatial information and/or with the information such as the overlapping status information of other tracks；Attribute information can be with Including：The time of occurrence of target object, the direction of motion of target object, license plate number, vehicle, the brand of vehicle, the color of vehicle, The dressing color of people, the age of people, the height of people, people whether wearing spectacles and/or people whether the information such as knapsack handbag.Target is former Figure can be the original image of video frame, or pass through the original image phase of mask figure and video frame and the image that obtains later.

It should be noted that target retrieval condition can be target object attribute information in arbitrary group of all information It closes, can also be some period.Established database is retrieved, is retrieved from database and meets target retrieval The track of the target object of condition.

Optionally, the target retrieval condition includes：The searching attribute information of retrieval time section and/or target object.

Specifically, when search condition only includes retrieval time section, the step that established database is retrieved Suddenly, may include：

According to retrieval time section, database is retrieved, obtains the track of all target objects in retrieval time section.

It should be noted that when target retrieval condition is some retrieval time section, obtain in retrieval time section The track of all target objects, for example, there are one section 7:00 to 12:00 when a length of 5 hours video, set target retrieval Condition is retrieval time section：8:00 to 9:00, then by 8:00 to 9:The track of all target objects extracts in 00, this implementation Example, is equivalent to and has carried out the interception of period to original video, can extract regarding for the movable more frequent period of target object Frequently, the input data that subsequent video frequency abstract generates is reduced, operand is reduced.

Specifically, when search condition only includes the searching attribute information of target object, it is described to established database The step of being retrieved may include：

According to the searching attribute information of target object, database is retrieved, is obtained and searching attribute information matches All tracks of all target objects.

It should be noted that target retrieval condition be target object attribute information in all information arbitrary combination When, established database is retrieved, the rail of target object identical with target retrieval condition is retrieved from database Mark.For example, target retrieval condition is：Between 1.75 meters of height, 40 years old to 45 years old age, the men of white down jackets is worn, then According to the target retrieval condition, the institute's rail for the target object for meeting the target retrieval condition can be extracted from database Mark.The present embodiment, target retrieval condition limit the target object in video, can ensure the accuracy of target object, It is easier to the track for the target object that extraction is met the requirements.

Specifically, when search condition includes the searching attribute information of retrieval time section and target object, it is described to built The step of vertical database is retrieved may include：

According to the searching attribute information of retrieval time section and searched targets object, database is retrieved, is retrieved In period, all tracks with the matched all target objects of searching attribute.

It is understood that in conjunction with above-mentioned two embodiment, the present embodiment can be got in retrieval time section and attribute All tracks of the target object of information matches, the present embodiment not only ensure that the active degree of target object but also defined target pair The attribute of elephant, compared to above-mentioned two embodiment, the track acquired is more accurate.

It is emphasized that target retrieval condition can be preset, that is, before obtaining video just The case where condition of target retrieval is set, is chiefly used in fixed scene, such as the same period of continuous many days daily Repeat some target object in same place, the case where for fixed scene, uses preset target retrieval condition It can be to avoid repeatedly setting same target search condition；Target retrieval condition can also be what user inputted according to actual conditions, Such as user needs to retrieve the track of specific objective object in some special time period, user can be according to the category of the target object Property information setting target retrieval condition.This is all reasonable.

It should be noted that when being retrieved to established database, can be generated according to target retrieval condition SQL (Structured Query Language structured query language) query statement；Established database is examined Rope.Wherein, SQL query statement is sentence the most commonly used in SQL query language, and SQL query language is a kind of data base querying And programming language, for accessing data and inquiry, update and management database；SQL query statement is by selecting to order It enables, the data that selection meets target retrieval condition in the table of database, which can be the track identification of target object Symbol, such as temporal information, the spatial information trace information of the track of target object can be determined by trajectory identifier.

S102 occurs overlapping at least two tracks according to overlapping status information, during the first track is gathered and is divided into one Group, every group is determined as a combined trajectories.

Wherein, overlapping status information can be the identifier for the track that overlapping target object occurs, and identifier can be The title of track, can also be track number or other are used to characterize the characteristic symbol of track；First track collection is combined into from number Combined by a plurality of track according to the target object for meeting target retrieval condition extracted in library at a track gather.

It should be noted that in order to improve the effect of the video frequency abstract in the case of track complexity of target object, Overlapping track will occur in the step of translating track along time shaft to hand over as a whole, in the set of the first track of combination Folded track obtains combined trajectories.Furthermore, it is desirable to which, it is emphasized that the sequence of each track on a timeline needs to protect in combined trajectories It holds constant.

S103 will meet the trajectory-offset of presetting translation condition to the same object time along time shaft in track to be translated Section.

Wherein, track to be translated includes：The combination rail of overlapping at least two tracks composition occurs in the set of first track Overlapping track does not occur in mark and/or the first track set；Target time section can be that user is preset, can also be Certain certain time in the reproduction time of entire video.It should be noted that the video frequency abstract to be generated is not simply It is spliced by motion segments, but the trajectory-offset of the target object by there are different time sections is to the same period, Concentrate the video frequency abstract formed.It is emphasized that be the translation to the time of track to the translation of the track of target object, and The translation of spatial position is not included.

S104, from the corresponding all first object artworks in all tracks obtained in database in target time section.

It should be noted that target artwork can be the original image of video frame, or pass through mask figure and video frame Original image phase and the image that obtains later, first object artwork be the target artwork of any track, be in first object artwork When the original image of video frame, first object artwork can be obtained according to the trace information of target object, can also be according to target pair The attribute information of elephant obtains.

S105 obtains the first abstract Background for generating video frequency abstract.

It should be noted that in video frame other than the first object artwork of target object, other content is formed by Image is the Background of video, and the video frequency abstract ultimately produced cannot include only first object artwork, should also be plucked including first Want Background.

It is emphasized that in the present embodiment, the first abstract Background can determine that method calculates by static background Obtained static abstract Background can also be the dynamic abstract Background that method determination is determined by dynamic background. When special screne or too long video time, video has prodigious difference in the Background of different time, is plucked to reduce video When generating, influence of the Background to video frequency abstract promotes the effect of video frequency abstract, needs to preserve video in different time periods Background is as abstract Background.

All first object artworks and the first abstract Background are spliced, generate video frequency abstract by S106.

Can be by first object it should be noted that first object artwork and the first abstract Background are spliced Artwork copies to the location of target object in the first abstract Background.But overlapped due to the track presence after translation Track and not overlapping track then simply cannot replicate pasting obtaining video frequency abstract to first object artwork.

Using the present embodiment, by being retrieved to established database, the track for meeting target retrieval condition is obtained, It is combined trajectories that overlapping track combination will occur in these tracks, then translates combined trajectories in different time periods and does not occur Overlapping track is to same target time section, finally to the corresponding target artwork in track and abstract Background in target time section Spliced, generates video frequency abstract；When generating video frequency abstract overlapping multiple target objects will occur for the embodiment of the present invention Track combination avoids losing in translation certain in overlapping track at a combined trajectories, on a timeline integral translation The visual effect for generating video frequency abstract is improved in track.

Optionally, the generation method of the video frequency abstract further includes：

When receiving interrupt requests input by user, terminate the flow that video frequency abstract generates；Receiving user's input Startup request when, execute obtain target retrieval condition, the step of retrieval to established database.

It should be noted that in order in the generating process of video frequency abstract, the usage experience of user is improved, user can be Random time inputs interrupt requests, for example, user has found target retrieval condition setting mistake, after receiving the interrupt requests, Terminate the flow of video frequency abstract generation；Then, user can reset target retrieval condition, and input and open according to demand Dynamic request again retrieves established database according to target retrieval condition after receiving startup request.

As shown in Fig. 2, a kind of generation method for video frequency abstract that the present embodiment is provided, the acquisition target retrieval condition The step of before, the generation method of video frequency abstract can also include：

S201 extracts all target objects from the video of input.

Wherein, target object is the target for having characteristic information, such as personage, automobile, steamer etc..

S202 extracts the trace information and attribute information of each target object.

Wherein, may include in trace information：The identifier of the track of target object, the tracing point quantity of target object, The frame number of each tracing point of target object, the temporal information of the track of target object, target object track space letter Breath and/or with the information such as the overlapping status information of other tracks；May include in attribute information：The time of occurrence of target object, The direction of motion of target object, license plate number, vehicle, the brand of vehicle, the color of vehicle, the dressing color of people, the age of people, people Height, people whether wearing spectacles and/or people whether the information such as knapsack handbag.

It should be noted that after getting the video of input, need to carry out video structural Objective extraction to the video, Video structural Objective extraction includes target object extraction and objective attribute target attribute extraction, is extracted by target object, obtains target track Mark describes file；It is extracted by objective attribute target attribute, obtains objective attribute target attribute and describe file, target trajectory describes file and retouched with objective attribute target attribute File is stated to be contained in video structural goal description file.Target object is extracted synchronous can carry out with objective attribute target attribute extraction, Objective attribute target attribute extraction is to extract one or multiple video frame images for including target object by default video frame extraction method, In conjunction with attributive classification device, the objective attribute target attribute that the result of synthesized attribute grader obtains target object describes file.Wherein, attribute The a certain generic attribute of grader target object for identification, and the generic can be obtained by the internal analysis of attributive classification device The information of property.Target object extraction is the particular community and movement properties of the target object extracted according to objective attribute target attribute, knot The particular community and movement properties for closing target object, and carry out multiple target tracking, by the particular community of the target object of tracking with Movement properties are associated fusion, and the target trajectory for obtaining target object describes file.Ensure in multiple targets while when occurring The track of target object and target object corresponds, and prevents the track of another target object from influencing the rail of this target object Mark improves the accuracy of target object extraction.

S203 stores the trace information of each target object and attribute information to video structural goal description file In.

It should be noted that video structural goal description file is used to store the attribute information and track letter of target object Breath.The generation of existing video structural goal description file is typically to be realized on industrial personal computer or server, certainly It can be realized by embedded platform, such as DSP (Digital Signal Processor, digital signal processor), ARM (Advanced Reduced Instruction Set Computer Machines, reduced instruction set computer microprocessor).

S204 generates database according to video structural goal description file.

It should be noted that after obtaining video structural goal description file, number is established using attribute information therein According to library, and by all attribute informations of the data base administration, including：The time of occurrence of target object, the movement side of target object To, whether the brand of license plate number, vehicle, vehicle, the color of vehicle, the dressing color of people, the age of people, the height of people, people wear Wear glasses and/or people whether the information such as knapsack handbag.

It should be noted that the process of Database and the prior art are essentially identical, distinguish only due to establishing Before database, when being analyzed, extracted the trace information of target object to the track of target object, it is extracted the friendship of target object Overlapping state information saves the overlapping status information of target object in database.

S205, from the video frame comprising target object, the artwork and mask figure of the corresponding each frame of extraction trace information, The target artwork of each frame corresponding with trace information is determined according to artwork and mask figure.

It should be noted that target artwork is the image of target object, in the present embodiment, target artwork can be by covering Code figure and the artwork phase of each video frame with obtain, since mask figure embodies the profile of target object, mask figure it is merely meant that The profile of target object, and do not include picture material, with the artwork phase of video frame with later to get to the region of mask figure The image of target object is more accurate compared to the image for extracting target object directly from video frame artwork.Wherein, mask figure It is to be obtained according to the motion mask of target object, the extractive technique of mask figure belongs to the prior art, and which is not described herein again.

As shown in figure 3, from regarding comprising target object described in a kind of generation method of video frequency abstract of the embodiment of the present invention In frequency frame, the step of the mask figure of the corresponding each frame of extraction trace information, may include：

S2051 extracts the motion mask of target object from the video frame comprising target object.

S2052 determines initial mask figure according to motion mask.

Wherein, motion mask is the 2-D data for constituting mask figure, by extracting the motion mask of target object, according to fortune The 2-D data of dynamic mask can determine the mask figure of target object.Mask figure characterizes the profile of target object, for that will regard Target artwork in frequency frame artwork is distinguished with Background.

S2053 determines the edge point set of initial mask figure.

S2054, the convex set that extraction marginal point is concentrated, constitutes the convex closure point set of mask figure.

S2055, the corresponding convex closure of filling convex closure point set, obtains final mask figure.

It should be noted that under complicated scene, easily there is mask figure and extract incomplete situation, in order to improve mask Figure, mask figure can be post-processed after extracting the mask figure of target object by further improving the effect of video frequency abstract Operation constitutes the convex closure point set of mask figure that is, according to the edge point set of mask figure, and fills the corresponding convex closure of convex closure point set, from And mask figure is improved, the profile of target object is embodied to the greatest extent.

S206 stores artwork, mask figure and target artwork into the database.

It should be noted that the corresponding artwork of target object, mask figure and target artwork are stored into database, so as to In the step of carrying out video frequency abstract generation, correspondence can be quickly found from database according to the attribute information of target object Target artwork.

Optionally, before artwork, mask figure and target artwork are stored the step into database, the life of video frequency abstract Can also include at method：

First, by predetermined period, abstract Background is obtained.

Wherein, predetermined period is to preserve the period of abstract Background, can be that user sets according to actual demand, also may be used To be the preset empirical value of those skilled in the art.

It should be noted that by the artwork phase of mask figure and video frame and target artwork can be obtained, due to mask figure Embody the profile of target object, mask figure it is merely meant that target object profile, and do not include picture material, with video frame Artwork phase with later to get to the image of the target object in the region of mask figure, compared to directly being extracted from video frame artwork The image of target object is more accurate, and under the premise of ensureing that target object extraction is complete, eliminates the back of the body in target frame Scape part promotes target artwork and abstract Background splicing effect, wherein mask figure is obtained according to the motion mask of target object , the extractive technique of mask figure belongs to the prior art, and which is not described herein again.During video analysis, from beginning to end all A Background is maintained, this Background is all updated per frame, and when reaching the predetermined period time, Background automatically saves one It is secondary.The update method of this Background is：To every frame image zooming-out motion mask figure, pixel is sport foreground, then the corresponding back of the body Otherwise the pixel of scape figure updates the pixel of Background, wherein more new formula is without update according to more new formula：A=b × k + c × (1-k), a are the Background pixel value of present frame, and b is the Background pixel value of former frame, and k is preset value, the value model of k The arbitrary number for 0 to 1 is enclosed, c is current video frame pixel value.

Secondly, the abstract Background in each period of acquisition is stored into the database.

It should be noted that since video is during broadcasting, Background can be varied from, but compared to target pair As the change very little of Background, therefore it may only be necessary to periodically preserve Background, not only ensure that the authenticity of background but also do not increase Add too many calculation amount.

As shown in figure 4, a kind of generation method of video frequency abstract of the embodiment of the present invention, the acquisition is plucked for generating video The step of the first abstract Background wanted, may include：

Target time section is divided into corresponding time subsegment by S1051 by the time corresponding to each predetermined period.

It should be noted that when periodically preserving Background, target time section is divided into corresponding time subsegment, it can To ensure, in the follow-up abstract Background obtained for generating video frequency abstract, can centainly get corresponding abstract background Figure determines Background of more preferably making a summary without more algorithms, can effectively save operand；Wherein, the time Subsegment corresponds to a predetermined period.

S1052 is determined in target time section, comprising corresponding first predetermined period of the most time subsegment in track.

S1053 obtains the corresponding first abstract Background of the first predetermined period from the database.

It should be noted that if the trace number for including in some time subsegment is most, illustrate in the period mesh The movement for marking object is most frequent, then the Background in the period should be used as the abstract Background of video frequency abstract, only portion less Point of rail mark is not inconsistent with real background, and static background figure compared to the prior art more can really embody the practical rail of target object Mark promotes the effect of video frequency abstract.

Using the present embodiment, by being retrieved to established database, the track for meeting target retrieval condition is obtained, It is combined trajectories that overlapping track combination will occur in these tracks, then translates combined trajectories in different time periods and does not occur Overlapping track is to same target time section, finally to the corresponding target artwork in track and abstract Background in target time section Spliced, generates video frequency abstract；When generating video frequency abstract overlapping multiple target objects will occur for the embodiment of the present invention Track combination avoids losing in translation certain in overlapping track at a combined trajectories, on a timeline integral translation The visual effect for generating video frequency abstract is improved in track.And by establishing database, preserve attribute information, the target of target object The information such as artwork improve the speed for generating video frequency abstract when generating video frequency abstract；Target artwork passes through mask figure and video frame Artwork phase with obtain, since mask figure embodies the profile of target object, mask figure it is merely meant that target object profile, without Including picture material, with video frame artwork phase with later to get to the image of the target object in the region of mask figure, compared to The image that target object is extracted directly from video frame artwork is more accurate.

As shown in figure 5, a kind of generation method of video frequency abstract of the embodiment of the present invention, it is described will rail be translated along time shaft The step of meeting trajectory-offset to the same target time section for presetting translation condition in mark, may include：

S1031 establishes queue to be translated and abstract queue.

Wherein, queue to be translated is the queue for storing target trajectory of not arranging, can be stored in queue to be translated Be that can also be there are no the track for judging whether to meet preset condition in database in the set of the first track not in target Between track in section.Abstract queue is the queue for storing the track for generating video frequency abstract.

S1032, track not in target time section is as track to be translated during the first track is gathered, and stores to waiting for It translates in queue.

Wherein, track to be translated be the first track set in not in the combined trajectories of target time section and do not overlap Track.

S1033, the track in target time section during the first track is gathered are stored into the abstract queue.

S1034 extracts current track to be translated from queue to be translated, and according to the artwork in database, obtains successively The rectangle frame of all target objects in the corresponding video frame in the current track to be translated.

It should be noted that present position is actually contained in a rectangle subgraph target object in the video frame, rectangle The size of subgraph is related with target extraction method, and the selected of rectangle subgraph belongs to the prior art, and which is not described herein again.Target object Track formed by multiple continuous video frame, then the track of a target object forms there are multiple rectangle subgraphs and waits translating The rectangle frame of track, the rectangle frame are all rectangle subgraphs comprising target object and the box of area minimum.

S1035, calculate the rectangle frames of all target objects respectively with stored to every rail of the abstract queue Overlapping area in the corresponding video frame of mark between the rectangle frame of target object.

S1036, when overlapping area is less than or equal to default Overlapping parameters threshold value, by current trajectory-offset to be translated to mesh The period is marked, and is stored to abstract queue.

It should be noted that when the overlapping area of the rectangle frame of two tracks is too big, illustrate the overlapping of this two tracks Part is too many, and it is therefore identical strip path curve when translating track, does not translate rectangle frame that can be approximately considered two tracks The too big track of overlapping area.In the present embodiment, an Overlapping parameters are preset, which can root It is set according to the specific object of actual conditions of demand and target object, is more than the default weight in the overlapping area of rectangle frame When folded parameter threshold, then it is assumed that lap is too many, does not store the track to queue of making a summary；Only in the overlapping area of rectangle frame When less than or equal to default Overlapping parameters threshold value, track to be translated just is stored to queue of making a summary.

It should be noted that the present embodiment by establish queue to be translated and abstract two queues of queue, by the first track Track in set not in target time section is deposited to queue to be translated, the rail during the first track is gathered in target time section Mark is deposited to abstract queue, then will be met the track for presetting overlay condition in queue to be translated and deposited to abstract queue.Using this implementation Example, can be to avoid when translating track, there is a phenomenon where entanglements.

Specifically, the corresponding all first object artworks in all tracks that target time section is obtained from database Step may include：

Obtain the corresponding all first object artworks in all tracks in abstract queue.

It should be noted that the track of trace information and target object and target object all has correspondence, certainly, The target artwork of trace information and target object also has correspondence.It therefore, can be according to the trace information of track to be translated The corresponding first object artwork of the trace information is extracted from database.

Optionally, described to splice all first object artworks and the first abstract Background, it generates video and plucks It wants, including：

First, according to the target frame information aggregate of target object in trace information, determine that first object artwork is plucked first Want the first position in Background.

It should be noted that further including in the trace information of every track：The target frame information aggregate of target object, target Frame information includes the coordinate and width height of the upper left angle point of the rectangle frame of target object, for example, target frame information is (x, y, w, h), Wherein, x is the abscissa of the upper left angle point of the rectangle frame of target object, and y is the vertical of the upper left angle point of the rectangle frame of target object Coordinate, w are the width of the rectangle frame of target object, and h is the height of the rectangle frame of target object, and certainly, target frame information can also wrap Center point coordinate and the width for including the rectangle frame of target object are high, for example, target frame information is (m, n, p, q), wherein m is target The abscissa of the central point of the rectangle frame of object, n are the ordinate of the central point of the rectangle frame of target object, and p is target object Rectangle frame width, q be target object rectangle frame height.Target forms the set of target frame information in moving process, should Set contains the information such as coordinate, length, direction of the first object artwork in the first abstract Background therefore can basis The target frame information aggregate of target object in trace information determines first of first object artwork in the first abstract Background It sets.

Again, each first object artwork is copied into corresponding first position in the first abstract Background, generates video Abstract.

It should be noted that when carrying out target artwork and the splicing for Background of making a summary, if directly by target artwork and Abstract Background is spliced, since method used in splicing may be such that position of the target artwork in Background of making a summary With physical location there are including part Background in error or target artwork, differed with Background to be spliced together, It is not inconsistent with artwork so as to cause position of the target artwork in Background of making a summary；Therefore in the present embodiment, target artwork can be The artwork of video frame and mask figure phase and obtained image, which does not include Background, and is embodied according to mask figure Go out actual position of the target object in Background of making a summary, it, can be accurately by mesh by target artwork and abstract Background splicing Mark artwork copies to the contour area that mask figure is sketched the contours, and can ensure the effect of target artwork and Background splicing of making a summary, from And improve the visual effect of video frequency abstract.

Specifically, the step that each first object artwork is copied to corresponding first position in the first abstract Background Suddenly, may include：

First, if target object has overlapping in each first object artwork, the overlapping part of corresponding track is set Pixel value be each target object target artwork pixel value mean value, the pixel value of overlapping part is not the mesh of each target object The pixel value for marking artwork, obtains figure to be copied；

Secondly, each first object artwork that each figure to be copied and target object do not overlap is copied into the first abstract Corresponding first position in Background generates video frequency abstract.

It should be noted that when generating video frequency abstract, the track presence for generating the target object of video frequency abstract is handed over It folds and not there is a situation where overlapping, when not occurring overlapping, directly target artwork can be copied in abstract Background, sent out When raw overlapping, according to the pixel value of lap, take the mean value of the pixel value of the target artwork of each target object as overlap The pixel value of the image divided, is then spliced again, generates video frequency abstract.Certain overlapping part can also take each target object The weighted value of the pixel value of target artwork, this is all reasonable.

As shown in fig. 6, a kind of generation method for video frequency abstract that the present embodiment is provided, in the acquisition target retrieval item Before the step of part, the generation method of video frequency abstract can also include：

S601 shows user interface according to user instruction.

It should be noted that user interface is to realize the interface interacted between user and system, user's interaction It can be dialog box in interface, or the selection picture in webpage.For prompting user to input target retrieval, default translation Condition and predetermined period, wherein predetermined period are for generating abstract Background.

S602 receives and preserves target retrieval condition, default translation item that user is inputted by the user interface Part and for generate abstract Background predetermined period.

Using the present embodiment, by being retrieved to established database, the track for meeting target retrieval condition is obtained, It is combined trajectories that overlapping track combination will occur in these tracks, then translates combined trajectories in different time periods and does not occur Overlapping track is to same target time section, finally to the corresponding target artwork in track and abstract Background in target time section Spliced, generates video frequency abstract；When generating video frequency abstract overlapping multiple target objects will occur for the embodiment of the present invention Track combination avoids losing in translation certain in overlapping track at a combined trajectories, on a timeline integral translation The visual effect for generating video frequency abstract is improved in track.And by user interface, user is supported to set target retrieval condition And arrange parameter, the flexibility of application is improved, is brought advantage to the user.

Optionally, the generation method of video frequency abstract can also include：

When receiving the startup request that user is inputted by the user interface, execute described to established number The step of being retrieved according to library；When receiving the interrupt requests that user is inputted by the user interface, terminate video The flow of summarization generation.

It should be noted that in order in the generating process of video frequency abstract, the usage experience of user is improved, user can be Random time inputs interrupt requests, for example, user has found target retrieval condition setting mistake, after receiving the interrupt requests, Terminate the flow of video frequency abstract generation；Then, user can reset target retrieval condition, and input and open according to demand Dynamic request again retrieves established database according to target retrieval condition after receiving startup request.

Corresponding to above method embodiment, an embodiment of the present invention provides a kind of generating means of video frequency abstract, such as Fig. 7 institutes It states, described device may include：

Retrieval module 710 retrieves established database for obtaining target retrieval condition, obtains comprising symbol Close the first track set of all tracks of the target retrieval condition, wherein be stored in the database from comprising target The trace information and target artwork for the every track extracted in the video frame of object wrap in the trace information of every track It includes：Overlapping status information between other tracks；

Composite module 720, for according to the overlapping status information, occurring to overlap extremely during first track is gathered Few two tracks are divided into one group, and every group is determined as a combined trajectories；

Translation module 730 presets the trajectory-offset of translation condition to same for that will meet in track to be translated along time shaft One target time section, wherein the track to be translated includes：It is not sent out in the combined trajectories and/or first track set Raw overlapping track；

First acquisition module 740 is corresponded to for obtaining all tracks in the target time section from the database All first object artworks；

Second acquisition module 750, for obtaining the first abstract Background for generating video frequency abstract；

Concatenation module 760, it is raw for splicing all first object artworks and the first abstract Background At video frequency abstract.

Using the present embodiment, by being retrieved to established database, the track for meeting target retrieval condition is obtained, It is combined trajectories that overlapping track combination will occur in these tracks, then translates combined trajectories in different time periods and does not occur Overlapping track is to same target time section, finally to the corresponding target artwork in track and abstract Background in target time section Spliced, generates video frequency abstract；When generating video frequency abstract overlapping multiple target objects will occur for the embodiment of the present invention Track combination avoids losing in translation certain in overlapping track at a combined trajectories, on a timeline integral translation The visual effect for generating video frequency abstract is improved in track.

Optionally, the target retrieval condition may include：The searching attribute letter of retrieval time section and/or target object Breath；

When search condition only includes retrieval time section, the retrieval module specifically can be used for：

According to retrieval time section, the database is retrieved, obtains all targets in the retrieval time section The track of object is as target trajectory；

When search condition only includes the searching attribute information of target object, the retrieval module specifically can be also used for：

According to the searching attribute information of the target object, the database is retrieved, obtains and belongs to the retrieval All tracks of all target objects of property information matches are as target trajectory；

When search condition includes the searching attribute information of retrieval time section and target object, the retrieval module, specifically It can be also used for：

According to the searching attribute information of retrieval time section and the searched targets object, the database is examined Rope obtains in the retrieval time section, and all tracks with the matched all target objects of the searching attribute are as target track Mark.

Further, comprising retrieval module 710, composite module 720, translation module 730, the first acquisition module 740, On the basis of second acquisition module 750, concatenation module 760, as shown in figure 8, a kind of video that the embodiment of the present invention is provided is plucked The generating means wanted can also include：

First extraction module 810, for extracting all target objects from the video of input；

Second extraction module 820, trace information and attribute information for extracting each target object, wherein believe track Breath includes：The mobile message of track and the overlapping status information between other tracks；

First memory module 830, for storing the trace information of each target object and attribute information to video Structured objects describe in file；

Database generation module 840, for according to the video structural goal description file, generating the database；

Third extraction module 850, for from the video frame comprising target object, extracting the corresponding each frame of trace information Artwork and mask figure, the target original of each frame corresponding with the trace information is determined according to the artwork and the mask figure Figure；

Second memory module 860, for storing the artwork, the mask figure and the target artwork to the data In library.

Using the present embodiment, by being retrieved to established database, the track for meeting target retrieval condition is obtained, It is combined trajectories that overlapping track combination will occur in these tracks, then translates combined trajectories in different time periods and does not occur Overlapping track is to same target time section, finally to the corresponding target artwork in track and abstract Background in target time section Spliced, generates video frequency abstract；When generating video frequency abstract overlapping multiple target objects will occur for the embodiment of the present invention Track combination avoids losing in translation certain in overlapping track at a combined trajectories, on a timeline integral translation The visual effect for generating video frequency abstract is improved in track.And by establishing database, preserve attribute information, the target of target object The information such as artwork improve the speed for generating video frequency abstract when generating video frequency abstract；Target artwork passes through mask figure and video frame Artwork phase with obtain, since mask figure embodies the profile of target object, mask figure it is merely meant that target object profile, without Including picture material, with video frame artwork phase with later to get to the image of the target object in the region of mask figure, compared to The image that target object is extracted directly from video frame artwork is more accurate.

As shown in figure 9, the third extraction module 850, may include：

First extracting sub-module 851, for from the video frame comprising target object, extracting the movement of the target object Mask；

First determination sub-module 852, for according to the motion mask, determining initial mask figure；

Second determination sub-module 853, the edge point set for determining the initial mask figure；

Second extracting sub-module 854, the convex set concentrated for extracting the marginal point, constitutes the convex closure point of the mask figure Collection；

Filling submodule 855 obtains final mask figure for filling the corresponding convex closure of the convex closure point set.

As shown in Figure 10, the translation module 730 may include：

Queue setting up submodule 731, for establishing queue to be translated and abstract queue；

First sub-module stored 732, for the not track in the target time section in gathering first track It as track to be translated, and stores into queue to be translated, wherein the track to be translated is in the set of first track The combined trajectories in the target time section and overlapping track does not occur；

Second sub-module stored 733, for the track in the target time section in gathering first track It stores into the abstract queue；

Third extracting sub-module 734, for the current track to be translated of extraction from the queue to be translated successively, and according to Artwork in the database obtains the rectangle frame of all target objects in the corresponding video frame in the track currently to be translated；

Operation submodule 735, for calculate the rectangle frames of all target objects respectively with stored to the abstract Overlapping area in the corresponding video frame in every track of queue between the rectangle frame of target object；

Third sub-module stored 736 is used for when the overlapping area is less than or equal to default Overlapping parameters threshold value, by institute Current trajectory-offset to be translated is stated to the target time section, and is stored to the abstract queue；

First acquisition module 740, specifically can be used for：

Obtain the corresponding all first object artworks in all tracks in the abstract queue.

Optionally, can also include in the trace information of every track：The target frame information collection of the target object It closes；

The concatenation module 760 may include：

Third determination sub-module, for the target frame information aggregate according to target object described in the trace information, really First position of the fixed first object artwork in the first abstract Background；

Video frequency abstract generates submodule, for each first object artwork to be copied to the first abstract Background In corresponding first position, generate video frequency abstract.

Optionally, the third determination sub-module, specifically can be used for：

If there is overlapping target object in each first object artwork, the pixel of the overlapping part of corresponding track is set Value is the mean value of the target artwork pixel value of each target object, the pixel value of overlapping part is not the target artwork of each target object Pixel value, obtain figure to be copied；

The concatenation module 760, specifically can be used for：

Each first object artwork that each figure and target object to be copied are not overlapped copies to described Corresponding first position in first abstract Background, generates video frequency abstract.

Optionally, described device can also include：

Computing module obtains abstract Background for pressing predetermined period；

Third memory module, for storing the abstract Background in each period obtained into the database；

As shown in figure 11, second acquisition module 750 may include：

Submodule 751 is divided, for by the time corresponding to each predetermined period, the target time section to be divided For corresponding time subsegment, wherein a time subsegment corresponds to a predetermined period；

4th determination sub-module 752 is corresponded to for determining in the target time section, comprising the most time subsegment in track The first predetermined period；

Background acquisition submodule 753, for from the database, obtaining first predetermined period corresponding first Abstract Background.

As shown in figure 12, described device can also include：

Display module 1210, for according to user instruction, showing user interface；

Receiving module 1220, the target retrieval item inputted by the user interface for receiving and preserving user Part, default translation condition and the predetermined period for generating abstract Background.

Optionally, shown device can also include：

Execution module, for when receiving the startup request that user is inputted by the user interface, executing institute State the step of being retrieved to established database；

Terminate module, for when receiving the interrupt requests that user is inputted by the user interface, terminating to regard The flow of frequency summarization generation.

It is understood that the generating means of video frequency abstract can wrap simultaneously in another embodiment of the embodiment of the present invention It includes：Retrieve module 710, composite module 720, translation module 730, the first acquisition module 740, the second acquisition module 750, splicing mould Block 760, the first extraction module 810, the second extraction module 820, the first memory module 830, database generation module 840, third Extraction module 850, the second memory module 860, computing module, third memory module, display module 1210, receiving module 1220, Execution module and terminate module.

It should be noted that herein, relational terms such as first and second and the like are used merely to a reality Body or operation are distinguished with another entity or operation, are deposited without necessarily requiring or implying between these entities or operation In any actual relationship or order or sequence.Moreover, the terms "include", "comprise" or its any other variant are intended to Non-exclusive inclusion, so that the process, method, article or equipment including a series of elements is not only wanted including those Element, but also include other elements that are not explicitly listed, or further include for this process, method, article or equipment Intrinsic element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that There is also other identical elements in process, method, article or equipment including the element.

Each embodiment in this specification is all made of relevant mode and describes, identical similar portion between each embodiment Point just to refer each other, and each embodiment focuses on the differences from other embodiments.Especially for system reality For applying example, since it is substantially similar to the method embodiment, so description is fairly simple, related place is referring to embodiment of the method Part explanation.

The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the scope of the present invention.It is all Any modification, equivalent replacement, improvement and so within the spirit and principles in the present invention, are all contained in protection scope of the present invention It is interior.

Claims (18)

1. a kind of generation method of video frequency abstract, which is characterized in that the method includes：

Target retrieval condition is obtained, established database is retrieved, obtains including to meet the target retrieval condition First track of all tracks gather, wherein be stored in the database and extracted from the video frame comprising target object The trace information and target artwork of every track, the trace information of every track include：It is overlapping between other tracks Status information；

Overlapping at least two tracks occur according to the overlapping status information, during first track is gathered and are divided into one Group, every group is determined as a combined trajectories；

The trajectory-offset of presetting translation condition will be met in track to be translated to same target time section along time shaft, wherein institute Stating track to be translated includes：Overlapping track does not occur in the combined trajectories and/or first track set；

The corresponding all first object artworks in all tracks in the target time section are obtained from the database；

Obtain the first abstract Background for generating video frequency abstract；

All first object artworks and the first abstract Background are spliced, video frequency abstract is generated.

2. the generation method of video frequency abstract according to claim 1, which is characterized in that the target retrieval condition includes： The searching attribute information of retrieval time section and/or target object；

It is described that established database is retrieved when search condition only includes retrieval time section, including：

According to retrieval time section, the database is retrieved, obtains all target objects in the retrieval time section Track；

When search condition only includes the searching attribute information of target object, described that established database is retrieved, packet It includes：

According to the searching attribute information of the target object, the database is retrieved, obtains and believes with the searching attribute Cease all tracks of matched all target objects；

When search condition include retrieval time section and target object searching attribute information when, it is described to established database into Row retrieval, including：

According to the searching attribute information of retrieval time section and the searched targets object, the database is retrieved, It obtains in the retrieval time section, all tracks with the matched all target objects of the searching attribute.

3. the generation method of video frequency abstract according to claim 1, which is characterized in that in the acquisition target retrieval condition Before, the method further includes：

All target objects are extracted from the video of input；

Extract the trace information and attribute information of each target object, wherein trace information includes：The mobile message of track, And the overlapping status information between other tracks；

The trace information of each target object and attribute information are stored into video structural goal description file；

According to the video structural goal description file, the database is generated；

From the video frame comprising target object, the artwork and mask figure of the corresponding each frame of extraction trace information, according to described Artwork and the mask figure determine the target artwork of each frame corresponding with the trace information；

The artwork, the mask figure and the target artwork are stored into the database.

4. the generation method of video frequency abstract according to claim 3, which is characterized in that described from regarding comprising target object In frequency frame, the mask figure of the corresponding each frame of extraction trace information, including：

From the video frame comprising target object, the motion mask of the target object is extracted；

According to the motion mask, initial mask figure is determined；

Determine the edge point set of the initial mask figure；

The convex set that the marginal point is concentrated is extracted, the convex closure point set of the mask figure is constituted；

The corresponding convex closure of the convex closure point set is filled, final mask figure is obtained.

5. the generation method of video frequency abstract according to claim 3, which is characterized in that it is described will rail be translated along time shaft Meet the trajectory-offset for presetting translation condition in mark to same target time section, including：

Establish queue to be translated and abstract queue；

The track in the target time section and is not stored to waiting putting down as track to be translated during first track is gathered It moves in queue, wherein the track to be translated is the not combination in the target time section in the set of first track Track and overlapping track does not occur；

The track in the target time section during first track is gathered is stored into the abstract queue；

Current track to be translated is extracted from the queue to be translated successively, and according to the artwork in the database, obtains institute State the rectangle frame of all target objects in the corresponding video frame in current track to be translated；

Calculate the rectangle frames of all target objects respectively with stored that every track of queue is corresponding regards to the abstract Overlapping area in frequency frame between the rectangle frame of target object；

When the overlapping area is less than or equal to default Overlapping parameters threshold value, by the current trajectory-offset to be translated to described Target time section, and store to the abstract queue；

The step of the corresponding all first object artworks in all tracks for obtaining the target time section from the database Suddenly, including：

Obtain the corresponding all first object artworks in all tracks in the abstract queue.

6. the generation method of video frequency abstract according to claim 3, which is characterized in that the trace information of every track In further include：The target frame information aggregate of the target object；

It is described to splice all first object artworks and the first abstract Background, video frequency abstract is generated, including：

According to the target frame information aggregate of target object described in the trace information, determine the first object artwork described First position in first abstract Background；

Each first object artwork is copied into corresponding first position in the first abstract Background, video is generated and plucks It wants.

7. the generation method of video frequency abstract according to claim 6, which is characterized in that described by each first object Artwork copies to corresponding first position in the first abstract Background, including：

If there is overlapping target object in each first object artwork, the pixel value that the overlapping part of corresponding track is arranged is The pixel value of the mean value of the target artwork pixel value of each target object, not overlapping part is the picture of the target artwork of each target object Element value, obtains figure to be copied；

Described that each first object artwork is copied to corresponding first position in the first abstract Background, generation regards Frequency is made a summary, including：

Each first object artwork that each figure and target object to be copied do not overlap is copied into first abstract Corresponding first position in Background generates video frequency abstract.

8. the generation method of video frequency abstract according to claim 3, which is characterized in that by the artwork, the mask Before figure and the target artwork are stored into the database, the method further includes：

By predetermined period, abstract Background is obtained；

The abstract Background in each period of acquisition is stored into the database；

The first abstract Background obtained for generating video frequency abstract, including：

By the time corresponding to each predetermined period, the target time section is divided into corresponding time subsegment, wherein One time subsegment corresponds to a predetermined period；

It determines in the target time section, comprising corresponding first predetermined period of the most time subsegment in track；

From the database, the corresponding first abstract Background of first predetermined period is obtained.

9. the generation method of video frequency abstract according to claim 8, which is characterized in that in the acquisition target retrieval condition Before, the method further includes：

According to user instruction, user interface is shown；

Receive and preserve target retrieval condition, default translation condition that user inputted by the user interface and for giving birth to At the predetermined period of abstract Background；

The method further includes：

When receiving the startup request that user is inputted by the user interface, execute described to established database The step of being retrieved；

When receiving the interrupt requests that user is inputted by the user interface, terminate the flow that video frequency abstract generates.

10. a kind of generating means of video frequency abstract, which is characterized in that described device includes：

Retrieval module retrieves established database for obtaining target retrieval condition, obtains including to meet the mesh Mark the first track set of all tracks of search condition, wherein be stored in the database from regarding comprising target object The trace information and target artwork for the every track extracted in frequency frame, the trace information of every track include：With other Overlapping status information between track；

Composite module, for according to the overlapping status information, overlapping at least two to occur during first track is gathered Track is divided into one group, and every group is determined as a combined trajectories；

Translation module, for along time shaft will wait translate meet trajectory-offset to the same target for presetting translation condition in track when Between section, wherein the track to be translated includes：Do not occur in the combined trajectories and/or first track set overlapping Track；

First acquisition module, for obtaining all tracks in the target time section corresponding all from the database One target artwork；

Second acquisition module, for obtaining the first abstract Background for generating video frequency abstract；

Concatenation module generates video for splicing all first object artworks and the first abstract Background Abstract.

11. the generating means of video frequency abstract according to claim 10, which is characterized in that the target retrieval condition packet It includes：The searching attribute information of retrieval time section and/or target object；

When search condition only includes retrieval time section, the retrieval module is specifically used for：

According to retrieval time section, the database is retrieved, obtains all target objects in the retrieval time section Track as target trajectory；

When search condition only includes the searching attribute information of target object, the retrieval module is specifically used for：

According to the searching attribute information of the target object, the database is retrieved, obtains and believes with the searching attribute All tracks of matched all target objects are ceased as target trajectory；

When search condition includes the searching attribute information of retrieval time section and target object, the retrieval module is specifically used for：

According to the searching attribute information of retrieval time section and the searched targets object, the database is retrieved, It obtains in the retrieval time section, all tracks with the matched all target objects of the searching attribute are as target trajectory.

12. the generating means of video frequency abstract according to claim 10, which is characterized in that described device further includes：

First extraction module, for extracting all target objects from the video of input；

Second extraction module, trace information and attribute information for extracting each target object, wherein wrapped in trace information It includes：The mobile message of track and the overlapping status information between other tracks；

First memory module, for storing the trace information of each target object and attribute information to video structural mesh In mark description file；

Database generation module, for according to the video structural goal description file, generating the database；

Third extraction module, for from the video frame comprising target object, extracting the artwork of the corresponding each frame of trace information With mask figure, the target artwork of each frame corresponding with the trace information is determined according to the artwork and the mask figure；

Second memory module, for storing the artwork, the mask figure and the target artwork into the database.

13. the generating means of video frequency abstract according to claim 12, which is characterized in that the third extraction module, packet It includes：

First extracting sub-module, for from the video frame comprising target object, extracting the motion mask of the target object；

First determination sub-module, for according to the motion mask, determining initial mask figure；

Second determination sub-module, the edge point set for determining the initial mask figure；

Second extracting sub-module, the convex set concentrated for extracting the marginal point, constitutes the convex closure point set of the mask figure；

Filling submodule obtains final mask figure for filling the corresponding convex closure of the convex closure point set.

14. the generating means of video frequency abstract according to claim 10, which is characterized in that the translation module, including：

Queue setting up submodule, for establishing queue to be translated and abstract queue；

First sub-module stored, for during first track is gathered not the track in the target time section as waiting putting down Track is moved, and is stored into queue to be translated, wherein the track to be translated is in the set of first track not described Combined trajectories in target time section and overlapping track does not occur；

Second sub-module stored is stored for the track in the target time section in gathering first track to institute It states in abstract queue；

Third extracting sub-module, for extracting current track to be translated from the queue to be translated successively, and according to the number According to the artwork in library, the rectangle frame of all target objects in the corresponding video frame in the track currently to be translated is obtained；

Operation submodule, for calculate the rectangle frames of all target objects respectively with stored it is every to the abstract queue Overlapping area in the corresponding video frame in track between the rectangle frame of target object；

Third sub-module stored is used for when the overlapping area is less than or equal to default Overlapping parameters threshold value, will be described current Trajectory-offset to be translated is stored to the target time section to the abstract queue；

First acquisition module, is specifically used for：

Obtain the corresponding all first object artworks in all tracks in the abstract queue.

15. the generating means of video frequency abstract according to claim 12, which is characterized in that believe the track of every track Further include in breath：The target frame information aggregate of the target object；

The concatenation module, including：

Third determination sub-module determines institute for the target frame information aggregate according to target object described in the trace information State first position of the first object artwork in the first abstract Background；

Video frequency abstract generates submodule, right in the first abstract Background for copying to each first object artwork The first position answered generates video frequency abstract.

16. the generating means of video frequency abstract according to claim 15, which is characterized in that the third determination sub-module, It is specifically used for：

If there is overlapping target object in each first object artwork, the pixel value that the overlapping part of corresponding track is arranged is The pixel value of the mean value of the target artwork pixel value of each target object, not overlapping part is the picture of the target artwork of each target object Element value, obtains figure to be copied；

The concatenation module, is specifically used for：

Each first object artwork that each figure and target object to be copied are not overlapped copies to described first Corresponding first position in abstract Background, generates video frequency abstract.

17. the generating means of video frequency abstract according to claim 10, which is characterized in that described device further includes：

Computing module obtains abstract Background for pressing predetermined period；

Third memory module, for storing the abstract Background in each period obtained into the database；

Second acquisition module, including：

Submodule is divided, for by the time corresponding to each predetermined period, the target time section to be divided into correspondence Time subsegment, wherein time subsegment corresponds to a predetermined period；

4th determination sub-module, for determining in the target time section, comprising the most time subsegment corresponding first in track Predetermined period；

Background acquisition submodule, for from the database, obtaining the corresponding first abstract back of the body of first predetermined period Jing Tu.

18. the generating means of video frequency abstract according to claim 10, which is characterized in that described device further includes：

Display module, for according to user instruction, showing user interface；

Receiving module, for receiving and preserving target retrieval condition that user inputted by the user interface, default flat Shifting condition and for generate abstract Background predetermined period；

Execution module, for when receiving the startup request that user is inputted by the user interface, it to be described right to execute The step of established database is retrieved；

Terminate module is plucked for when receiving the interrupt requests that user is inputted by the user interface, terminating video The flow to be generated.